TerraBench
Explore score history ↗TerraBench is a benchmark for grounded Earth-science reasoning, built on TerraAgent, a ReAct-style framework that couples LLM planning with scientific tools. It includes 403 tasks across three tracks and eight domains with 24,500 verified steps.
- Released
- 2026-06-11
- Institutions
- Mohamed bin Zayed University of Artificial Intelligence
- Evaluates
- General AI, Information retrieval, Language & Knowledge, Planning, Reasoning
- Openness
- unknown
- Importer
- Claire Radar
- Review status
- unreviewed
TerraBench paper
No reported scores are on record for this benchmark yet.